Skip to content
@PSI-TAMU

PSI Lab @ Texas A&M

We study how artificial and biological sensory systems interact with and adapt to their environments using deep learning models.

PSI Lab

Perception, Sensing, and Instrumentation Laboratory
Department of Computer Science and Engineering · Texas A&M University

Machine learning for speech, sensing, and human health.

Lab Website · Publications · All Repositories


About Us

The Perception, Sensing, and Instrumentation (PSI) Lab at Texas A&M University, directed by Prof. Ricardo Gutierrez-Osuna, develops machine learning and signal processing methods for human-centered technologies.

Our work spans speech processing and digital health: from real-time, privacy-preserving speech systems to wearable sensing and data-driven approaches for understanding physiological signals. Across both areas, we study robust representations, efficient models, and intelligent systems that work beyond controlled laboratory settings.

Research Areas

Speech & Spoken Communication

We build models that analyze, transform, and generate speech while preserving the information that matters for communication.

  • Voice conversion and speech synthesis: Streaming and low-latency methods for controllable speech generation.
  • Speaker privacy: Anonymization techniques designed to protect speaker identity while preserving linguistic content.
  • Accent and pronunciation modeling: Accent classification, adaptation, and approaches supporting non-native speech and pronunciation learning.
  • Robust speech representations: Methods for disentangling speaker, linguistic, and prosodic information.

Digital Health & Physiological Sensing

We develop data resources and machine learning methods that use physiological signals to investigate health-related questions.

  • Wearable biosignal analysis: Learning from ECG, PPG, electrodermal activity, accelerometry, and other sensor streams.
  • Glucose monitoring and diabetes research: Exploring non-invasive estimation, hypoglycemia prediction, and multimodal physiological modeling.
  • Personalized nutrition: Understanding glucose responses to meals and other lifestyle factors.
  • Reliable evaluation: Data processing and assessment tools for physiological prediction models.

Our broader research also includes active sensing, chemical sensing, biofeedback, and assistive technologies.

Featured Open-Source Projects

Speech Processing

Project Description Resources
AccentCL Robust English accent classification with incremental expansion to new accent categories. SLT 2026. Paper · Demo
TVTSyn Content-synchronous, time-varying timbre modeling for streaming voice conversion and anonymization. ICLR 2026. Paper · Demo
DarkStream Low-latency, real-time speech anonymization with a streaming waveform-based model. ASRU 2025. Paper · Demo
streamVC An end-to-end streaming framework for efficient, low-latency speaker anonymization. SLT 2024. Paper · Demo

Digital Health

Project Description Resources
PhysioCGM A multimodal physiological dataset combining wearable biosignals with continuous glucose measurements for glucose-estimation research. Scientific Data, 2025. Paper · Data and code
CGMacros A dataset and analysis resources for studying personalized nutrition, meal composition, and glucose responses. Scientific Data, 2025. Paper · Code

Pinned Loading

  1. anony_benchmark anony_benchmark Public

    Python

  2. TVTSyn TVTSyn Public

    Official code repository for "TVTSyn: Content-Synchronous Time-Varying Timbre for Streaming Voice Conversion and Anonymization"

    Jupyter Notebook 5 2

  3. AccentCL AccentCL Public

    Official code for AccentCL: Robust Accent Classification with Incremental Expansion (SLT 2026)

    Jupyter Notebook 2

Repositories

Showing 10 of 11 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…